1. Scraping Data

Like we mentioned, the current dataset provided by Billboard does not have the top 100 songs for 2017 and 2018. Utilizing the rvest package to scrape the Wikipedia website we were able to grab the remaining songs’ titles and artists for the Billboard top 100!

url_2017 <- "https://en.wikipedia.org/wiki/Billboard_Year-End_Hot_100_singles_of_2017"
url_2018 <- "https://en.wikipedia.org/wiki/Billboard_Year-End_Hot_100_singles_of_2018"

top_100_2017 <- url_2017 %>%
  read_html() %>%
  html_nodes(xpath='//*[@id="mw-content-text"]/div/table') %>%
  html_table()
top_100_2017 <- top_100_2017[[1]]

top_100_2018 <- url_2018 %>%
  read_html() %>%
  html_nodes(xpath='//*[@id="mw-content-text"]/div/table') %>%
  html_table()
top_100_2018 <- top_100_2018[[1]]

top_100_2017['Year'] <- 2017
top_100_2018['Year'] <- 2018

colnames(top_100_2017) <- c("no", "title", "artist", "year")
colnames(top_100_2018) <- c("no", "title", "artist", "year")

top_100_2017[] <- lapply(top_100_2017, gsub, pattern='"', replacement='')
top_100_2018[] <- lapply(top_100_2018, gsub, pattern='"', replacement='')

top_100_2017$lyrics <- NA #set lyrics to NA (going to calculated later)
top_100_2018$lyrics <- NA

new_df  <- rbind(top_100_2017, top_100_2018)
df_all_no_lyrics <- rbind(ranked_data, new_df)

na_lyrics <- df_all_no_lyrics %>%
  filter(is.na(df_all_no_lyrics$lyrics)) #gets all Na lyrics
lyric_no_NA <- na.omit(df_all_no_lyrics) #saves all data that does not have NA

datalist <- list()
for (row in 1:nrow(na_lyrics)) { #loops through rows of dataset
  print(row) # print to follow code execution
  x <- possible_lyrics(artist = str_remove_all(na_lyrics[3]$artist[row], '"'), song = str_remove_all(na_lyrics[2]$title[row], '"'), info = "simple") # Finds lyrics
  if (grepl("featuring", na_lyrics[3]$artist[row])) { # remove featuring artists as it cant find the learics if it is included in artist
    z <- gsub("featuring.*","",na_lyrics[3]$artist[row])
    x <- possible_lyrics(artist = substr(z, 1, nchar(z)-1), song = str_remove_all(na_lyrics[2]$title[row], '"'), info = "simple")
    if (length(x) == 0) {
      datalist[[row]] <- NA
    }else {
      datalist[[row]] <- x %>%
          summarise(lyrics = paste(lyric, collapse=", "))
    }
  }else if (length(x) != 0){
    datalist[[row]] <- x %>%
        summarise(lyrics = paste(lyric, collapse=", "))
  }else {
    datalist[[row]] <- NA
  }
}
lyrics_data = do.call(rbind, datalist) # add all lyrics to datalist
na_lyrics$lyrics <- lyrics_data

lyrics_added <- rbind(lyric_no_NA, na_lyrics) # add the newly found lyrics with data with lyrics
# saveRDS(lyrics_added, file = "data_in/all_data_lyrics.rds")

all_data_lyrics <- readRDS("data_in/all_data_lyrics.rds")

artist_genre <- list()
for (row in 1:nrow(all_data_lyrics)) {
  print(row)
  if (row == 4627 | row == 4626 | row == 4628) { # certain entries does not want to run 
    next
  }
  if (grepl("featuring", all_data_lyrics[[3]][row])) { # deal with featuring artists
    
    z <- gsub("featuring.*","",all_data_lyrics[[3]][row])
    genre <- searchArtist(substr(z, 1, nchar(z)-1), token = my_oauth)[5]$genres[1] #gets genres
    
    if (is.null(genre)) {
      if (grepl("and", all_data_lyrics[[3]][row])) {
        
        x <- gsub(" and.*","",all_data_lyrics[[3]][row])
        x[] <- lapply(x, gsub, pattern='"', replacement='')
        genre <- searchArtist(x[[1]], token = my_oauth)[5]$genres[1] #gets genres
        
        if (is.null(genre)) {
          artist_genre[[row]] <- NA
        }else {
          artist_genre[[row]] <- sub("\\,.*", "", genre)
        }
        
      }else{
        artist_genre[[row]] <- NA
      }
    }else {
      artist_genre[[row]] <- sub("\\,.*", "", genre)
    }
    
  }else if (grepl("and", all_data_lyrics[[3]][row])) { # if artist contains 'and'
    
    x <- gsub(" and.*","",all_data_lyrics[[3]][row])
    x[] <- lapply(x, gsub, pattern='"', replacement='')
    genre <- searchArtist(x[[1]], token = my_oauth)[5]$genres[1] #gets genres
    if (is.null(genre)) {
      artist_genre[[row]] <- NA
    }else {
      artist_genre[[row]] <- sub("\\,.*", "", genre)
    }
    
  }else {
    
    genre <- searchArtist(all_data_lyrics[[3]][row],token = my_oauth)[5]$genres[1] #gets genres
    if (is.null(genre)) {
      artist_genre[[row]] <- NA
    }else {
      artist_genre[[row]] <- sub("\\,.*", "", genre)
    } 
    
  }
}
genre_data <- do.call(rbind, artist_genre)
data_with_genre <- cbind(lyrics_added, genre_data) # add all data together
# saveRDS(data_with_genre, 'all_data_genre_not_generalised.rds')

all_data_genre_ungen <- readRDS("all_data_genre_not_generalised.rds")

pop <- c('dance pop', 'pop', 'pop christmas', 'post-teen pop', 'new wave pop', 'brill building pop', 'bubblegum pop', 'canadian pop', 'boy band', 'europop', 'viral pop', 'synthpop', 'pop punk', 'canadian hip hop', 'afropop', 'classic girl group', 'girl group', 'australian dance', 'k-hop')

rap <- c('pop rap', 'rap', 'southern hip hop', 'hip hop', 'trap music', 'dirty south rap', 'gangster rap', 'east coast hip hop', 'hardcore hip hop', 'hip pop', 'hip house', 'rap metal', 'dwn trap', 'emo rap')

rock <- c('mellow gold', 'soft rock', 'rock', 'album rock', 'pop rock', 'classic rock', 'folk rock', 'hard rock', 'dance rock', 'roots rock', 'southern rock', 'blues-rock', 'alternative rock', 'funk rock', 'psychedelic rock', 'rock-and-roll', 'modern rock', 'art rock', 'classic funk rock', 'adult standards', 'rockabilly', 'merseybeat', 'surf music', 'deep adult standards', 'rhythm and blues', 'louisiana blues', 'blues', 'melodic metalcore', 'glam metal', 'alternative metal', 'comic metal', 'nu metal')

electronic <- c('tropical house', 'edm', 'brostep', 'bmore', 'broken beat', 'dark trap', 'deep groove house', 'retro electro', 'drum and bass', 'italian techno', 'electronic', 'dubstep', 'finnish electro', 'complextro', 'house', 'cyberpunk', 'electro')

r_and_b <- c('r&b', 'urban contemporary', 'deep pop r&b', 'indie r&b', 'doo-wop', 'modern reggae', 'easy listening', 'reggae', 'uk reggae')

soul  <- c('motown', 'disco', 'soul', 'funk', 'neo soul', 'soul christmas', 'new jack swing', 'memphis soul', 'chicago soul', 'southern soul', 'soul blues', 'post-disco', 'quiet storm', 'jazz blues', 'jazz funk', 'electric blues', 'latin', 'christian relaxative', 'memphis soul', 'classic soul', 'indie jazz', 'bebop', 'dixieland', 'avant-garde jazz', 'jazz trumpet', 'smooth jazz', 'classical')

country <- c('country',' contemporary country', 'country christmas', 'country road', 'modern country rock', 'country rock', 'traditional country', 'folk christmas', 'folk', 'folk-pop', 'nashville sound', 'country dawn', 'traditional folk', 'outlaw country', 'country gospel','arkansas country', 'american folk revival', 'alabama indie', 'london indie', 'alaska indie')

other <- c('beach music', 'background music', 'rif', 'disney', 'chillwave', 'comic', 'british invasion', 'compositional ambient', 'canadian celtic')

all_data <- all_data_genre_ungen

all_data$genre <- ifelse(grepl(paste(pop, collapse = "|"), all_data$genre_data), "pop",
             ifelse(grepl(paste(rap, collapse = "|"), all_data$genre_data), "rap",
             ifelse(grepl(paste(rock, collapse = "|"), all_data$genre_data), "rock",
             ifelse(grepl(paste(electronic, collapse = "|"), all_data$genre_data), "electronic",
             ifelse(grepl(paste(r_and_b, collapse = "|"), all_data$genre_data), "r&b",
             ifelse(grepl(paste(soul, collapse = "|"), all_data$genre_data), "soul",
             ifelse(grepl(paste(other, collapse = "|"), all_data$genre_data), "other",      
             ifelse(grepl(paste(country, collapse = "|"), all_data$genre_data), "country", "other"))))))))
all_data$genre_data <- NULL


# saveRDS(all_data, file = "data_in/all_data.rds")

all_data <- readRDS("data_in/all_data.rds")

2. Cleaning

Now that we had all our data updated, the next phase was to start cleaning to ensure an accurate representation.

Separating Artists One of the biggest challenges was seperating the artists from the featured artists. This posed an indepth cleaning regime by using regular expressions and packages from R to seperate the different artists into their own respective columns. By doing this, we can create a network of artists and their collaborators, and link their respective featured songs as edges. Another challenge was collecting lyrics for songs that were not available on the original dataset - so we had to pull it from an external source via a scraper.

Generalizing Genres As genres are an important aspect of this analysis - we had to pull in genres from a Spotify API, and generalise them. This took some time to sort them into generalizations and checking smaller samples to ensure the results made sense. This provides the opportunity the gain insight into which artists tend to collaborate together and which genre is the most popular among the community of artists.

Cleaning Lyrics Like a motherbird unnesting its tokens we use the unnest function to find the words who could fly at the top with the highest occurrence frequency. This allows us to find words that are uncommonly long or spelt so we can go back to the nest(corpus) and utilize the appropriate cleaning function.

all_stops <- c(stopwords::data_stopwords_snowball$en,stopwords::data_stopwords_smart$en,"the",'" The"',"told", "and.")
all_stops <- data.frame(all_stops)  #These are words that occur frequently but add no value to the term frequency and only clutter the results.
colnames(all_stops) <- "word"

Doing term frequency on the lyrics column and unnesting the words to find the most common terms and dirty data.

all_data <- data.frame(all_data)
Terms <- all_data %>%  #This chunk unnests the tokens in the dataset. These tokens are the single words in a sentence. It then counts the occurence of each word.
  unnest_tokens(word,lyrics) %>%
  anti_join(all_stops, by = 'word')%>%
  count(word, sort = TRUE)%>%
  mutate(len=nchar(word))

By utlizing term frequency we were able to detect common filthy data, like extra whitespace and lyric identifiers such as [Chorus] and [Verse].

all_data <- as.data.table(all_data) #Cleaning functions to remove non-text data and song discriptual data such as [Verse 1] and splitting words attached to commas
all_data$lyrics<-  gsub("([a-z])([A-Z])", "\\1 \\2", all_data$lyrics)

all_data[, lyrics := str_replace_all(lyrics, pattern = "\\[(.*?)\\]", "")]
all_data$lyrics<-  gsub("([a-z]) ([A-Z])", "\\1. \\2", all_data$lyrics)
all_data$lyrics<-  gsub(",", " ",all_data$lyrics)

# all_data_clean <- saveRDS(all_data,"data_out/all_data_clean.rds")

Clean data Saved

all_data_clean <- readRDS("data_out/all_data_clean.rds")

Question 1.1 : Most promonent words in lyrics

The most prominent words present in the entire corpus over the last 58 years. Word clouds are always a good tool to visualize all the facts. By unnesting the different words within the lyrics column and transforming it into a document term matrix we were able to create a dataset consisting of only the most prominently used words.As you can see, the heart wants what it wants. Love will always be on the top of our silly human minds as well as your significant other not - not 3 month old humans.

Question 2.1 :

Most promonent curse words in lyrics. What is a song without some explicity? The following word cloud displays the most prominent swearwords. The swearword dictionary was provided by yours truly.

swear_words <- c("fuck","shit","freak","bastard","sex","damn","ass","cock","penis","pussy","dick","lame","slut","whore","motherfucker","hell","prick","crap","fucking")
swear_words <- as.data.frame(swear_words)
swear_words <- swear_words%>%
  rename("word" = swear_words)

d_swear <- d


swear_terms <- d_swear%>%
  anti_join(stop_words, by = "word")%>%
  right_join(swear_words, by = "word")

The Word Cloud of all the most prominent swear words in the lyrics. Swearing is not cool kids. Stay in school.

Question 2.2 :

Phew, those were some hefty answers in the swear-word-cloud. Has music always been as vulgar as this? Looking back, using the oh-so-useful Spotify data we can identify if a song is generalized as explicit or not. Clearly those old people were on to something as the vulgar-ness of music has risen over the last phew decades. Being almost non existent. Peaking in the early 2000’s to an absurd degree of more than 30 explicit songs on the billboard 100. One can argue that since the dawn of the new decade a more modern PC culture has been adapted - for better and for worse - therefore resulting in the explicitness of the modern treffer to be tamer than its predecessors.

Question 3 : Most danceable era’s and tempo’s in most and least danceable songs

It seems there is always something about these hits that make them reach the infamous Hot 100. Maybe we feel our inner child come alive with every catchy “ohh-lala”, or hearing it on the radio 5 times a day makes them an integral part of our routines. But just how well do these big tracks fare on the dancefloor?

Luckily, the spotify_track_data dataset has worked the numbers for us on how danceable each song on the charts actually is. According to these numbers, the higher the danceabilty, the more we’ll be breaking it down and begging the DJ for more.

Firstly, we’ll look at how the danceability of tracks has changed throughout the years. In our regression graph below, little to our surprise, you can see that the disco era has proven in the numbers to have produced the music that speaks to our body and soul. Yes, boogie will always be king!

From then on, it seems we have to make peace with the dying days of dance, and accept that flossing to Justin Bieber’s Despacito is the new Saturday Night Fever(sorry Travolta). The regression line doesn’t lie, neither do our hips.

Now we’re sure your mind has wondered, what are THE ultimate dance treffers?

We’re here to ease your un-answered pains.

Again, the numbers say might bust-a-move, but maybe these don’t quite make your go-to party playlist.

Planning a braai-turned disturb-the-neighbours kind of night? Test-drive these top ten tuuunes, and swiftly avoid the bottom ten, and we can suggest which songs may fall into similiar categories.

Here one can see that the circle indicates the different years and the squares represent the dancibility. This shows us how which years are linked to which degree of dancibility.

What is the most likely element of a song that brings us to our feet other than it being written by Neil Diamond and titled as an agreeable popular female name? WHO KNOWS! But we’ll take a wild swing at Tempo for the time being.

When looking at the Tempo numbers, the most danceable songs have an average Tempo sitting around 116 beats per minute. Go too fast, the party can go from sweet sing-alongs to a gym class your never signed up for. Go too slow, you find yourself with sweaty and anxious thoughts of years in fairy-lit school halls, not being asked to dance, in a dress you payed waaay too much for, for this lame night anyway.

So mark our words - the bottom ten are a good reminder to keep it at a constant Tempo. (damn should we be charging for life advice too?)

The range of this graph goes from below 50 beats per minute to over 200 beats per minute.

This differs from the top danceable songs, where the graph’s danceablity numbers range from 100 to 120 beats per minute. The majority of the songs also float close to the mean line.

This shows that a tempo around 116-120 beats per mintue is the best for a popular dancing song.

Question 4 :

But who is crowned top of the Billboard Top 100? Let’s have a look at who is featured the most from 1960-2018.

From the plot above, you can clearly see who the top dog is. It seems that time does not go slowly for her, as Madonna has reached the top 100 over 30 times! This can be attributed to the fact that she was super popular during the 1980’s till the mid 2000’s, releasing smash hit after smash hit. All hail the original ‘Queen of Pop’! Following this, the next topic will investigate the sentiment of lyrics – let’s see what the most positive and most negative words were throughout the history of billboard bangers!

Question 5 :Sentiment Analysis

There’s a song for every mood, and maybe some a bit more morbid than others. Ever wondered about the lyrics of the songs you listen to? Upbeat tunes, but some pretty bleak lyrics, means that these songs will not score well on our sentiment analysis.

As an interesting topic, we decided to run sentiment analysis on the lyrics provided by each song and artist, to determine if artists over the years tend to write about more negative or positive sentiments. Sentiment dispersion is very diverse, as to be expected when dealing with a genre as diverse as music. Adding valence shifters to try and compensate for the probable words that can effect a negative/positve word does not add much to the overall sentiments as bands tend to repeat negative words like ‘Shake’ and that is according to Jockers/Rinkers. Bad. Old People amiright?

## Warning: Each time `sentiment_by` is run it has to do sentence boundary disambiguation when a
## raw `character` vector is passed to `text.var`. This may be costly of time and
## memory.  It is highly recommended that the user first runs the raw `character`
## vector through the `get_sentences` function.

## Warning: Removed 2 rows containing missing values (geom_point).

## Warning: Removed 2 rows containing missing values (geom_segment).

Question 6 : Can Community Detection on Collaboration Determine Genre?

To answer this question we followed the following method:

1. Extract all the artists who collaborated on a given song.
2. Repeat that for 5700 songs.
3. Do simple community detection on the data.
4. Map it in gephi and observe results.
5. Map accutal Genres on the community graph and work out ratios in R.
6. Determine if our question is valid.

Preparing data for question For step 1 we had to split the artist column where artist collaborated on a song. Collaborations were identified by the key words: ‘featuring’, ‘and’, ‘with’, ‘,’ and ‘&’. There are some artists that have these keywords, they were excluded for the extraction process.

The code we used to achieve this was as follows:

all_data <- readRDS("data_in/all_data.rds")
all_data$lyrics <- NULL
split_feat <- str_split_fixed(all_data$artist, 'featuring', 2) # split the rows based on 'featuring'
split_feat_song <- cbind(all_data, split_feat) # add songs to df
split_feat_song$artist <- NULL
colnames(split_feat_song) <- c('no', 'title', 'year', 'genre', 'artist_1', 'artist_2')

only_colab <- split_feat_song %>% # Remove the empty columns (no collaborations)
  filter(split_feat_song$artist_2 != "")

only_colab <- only_colab[c('artist_1','artist_2','genre','title','year', 'no')] # change Column order

only_colab <- separate_rows(only_colab, artist_1, sep = ', ') # split into rows on key word ',' for artist_1
only_colab <- separate_rows(only_colab, artist_2, sep = ', ') # split into rows on key word ',' for artist_2
only_colab <- separate_rows(only_colab, artist_2, sep = 'and ') # split into rows on key word 'and' for artist_2 (there are no artists that need 'and')

only_colab <- as.data.table(only_colab) 
only_colab[, artist_1 := str_remove_all(artist_1, pattern = "^\\s"),""] # Removes whitespace
only_colab[, artist_2 := str_remove_all(artist_2, pattern = "^\\s"),""] # Removes whitespace
only_colab[, artist_1 := str_remove_all(artist_1, pattern = "\\s$"),""] # Removes whitespace
only_colab[, artist_2 := str_remove_all(artist_2, pattern = "\\s$"),""] # Removes whitespace

only_colab_stay <- only_colab %>% # Artist that should not be split.
  filter(artist_1 == "Marky Mark and the Funky Bunch" |
         artist_1 == "Michael Franti & Spearhead" |
         artist_1 == "Timbaland & Magoo" |
         artist_2 == "Trina & Tamara" |
         artist_2 == "8Ball & MJG")

# Subset these artist to that they are not affected by the split
only_colab_rem <- subset(only_colab, artist_1 != "Marky Mark and the Funky Bunch")
only_colab_rem <- subset(only_colab_rem, artist_1 != "Michael Franti & Spearhead")
only_colab_rem <- subset(only_colab_rem, artist_1 != "Timbaland & Magoo")
only_colab_rem <- subset(only_colab_rem, artist_2 != "Trina & Tamara")
only_colab_rem <- subset(only_colab_rem, artist_2 != "8Ball & MJG")

# split remaining artist on key words
only_colab <- separate_rows(only_colab_rem, artist_1, sep = 'and ')
only_colab <- separate_rows(only_colab, artist_1, sep = '& ')
only_colab <- separate_rows(only_colab, artist_2, sep = '& ')
only_colab <- separate_rows(only_colab, artist_1, sep = 'with ')

# Artists that spelling differed. Thus, changed to make the names uniform
colab_all <- rbind(only_colab, only_colab_stay)
colab_all[620]$artist_2 <- "Tyga"
colab_all[621]$artist_2 <- "Jay-Z"
colab_all[577]$artist_1 <- "Jay-Z" 
colab_all[743]$artist_1 <- "Jay-Z" 
colab_all[745]$artist_1 <- "Jay-Z" 
colab_all[733]$artist_1 <- "Ty Dolla Sign"
colab_all[712]$artist_2 <- "Ty Dolla Sign"
colab_all[251]$artist_1 <- "The East Side Boyz"
colab_all[242]$artist_2 <- "The East Side Boyz"
colab_all[13]$artist_2 <- "Snoop Dogg"

colab_all <- colab_all[!(colab_all$artist_2 == ""), ] # Remove Null entries
# Check and remove if any more whitespace developed from the split
colab_all[, artist_1 := str_remove_all(artist_1, pattern = "^\\s"),""]
colab_all[, artist_2 := str_remove_all(artist_2, pattern = "^\\s"),""]

colab_all[, artist_1 := str_remove_all(artist_1, pattern = "\\s$"),""]
colab_all[, artist_2 := str_remove_all(artist_2, pattern = "\\s$"),""]

# saveRDS(colab_all, "data_in/colaboration_data.rds")

Now the data should be ready and all the collaborating artists will be split! Let’s find out what these connections between the collaborating artists mean.

Communities are an important part to the analysis of this dataset. Although genres act as a sort of label for artists, we want to find out if community detection can correlate with the labelled genres. By forming clusters with collaborating artist we are able to detect inherent communities via a community detection aldu-rhythm called Louvain.

It was assumed that collaborations would primarily occur in the same genre. We thus ran the Louvain community detection algorithm to distinguish the different artists in the defined communities. The reason why we choose the Louvain method is because of its fast processing time, responsiveness and accuracy on bigger datasets like this one.

This was the result after applying Louvain community detection:

The graph above was coloured according to different communities.
Originally there was 100 different identified communities but for visualization we filtered for the 9 biggest communities. The remaining 91 communities, however, were not exlcuded from our data analysis process. Upon observation, there definitely seems to be a distinction between where artists primarily collaborate. With many of them staying within their communities and a few branching out to surrounding communities.

So is there a correlation between collaborations and genres? Looking back at our question, we need to determine if this graph does in fact show genre. By just looking at graph, the it does seem that similar artists group together (like that of Drake, Kanye West and DJ Khaled ) making us positive about the results we’ll get back from the actual stats. To truly see wether these communities are showcasting genres we needed to extract the genre-ratios found in each community. This will show us wether one genre is promonent within a given community.

To do this we took the collaboration dataset and compared the edge & node tables determined:

1. Which genres where in a give community &
2. What were the genre-ratio proportinal to the community’s size.

Our findings: From the 100 identified communities 70% of them had one primary genre and 94% of the communities had a genre that made up more than 50% of the community’s size. This is great news as it shows how the community detection had a high level of accuracy in determining the genre communities. Below are a few samples from the Dataset and their respective genre ratios.

From this we can concluded that the factors that were leading groups to collaborate in communities could definitly be linked to the genres of those artists. Thus this form of community detection can help identify genre collaborations.

Question 7 : Are there opportunities for more collaboration through mutual ties?

A bridge is a type of social tie that connects two different groups in a social network. In this case, this could be where your favorite artist collaborates with another genre or group of artists. Making that interesting music we know you love to listen!

Highlighting the bridges between the different clusters helps us identify the way in which communities can be linked to each other. Bridges in our case are amazing opportunities for more mixes to our grooves. Who wants just some plain old country when you could be jamming out to possible canadian-country-punk-rap collaborations.

We were able to observe such an ocurrence in Drake’s pop community. As you can see by the weighting in the edges( which were weighted according to the amount of collaboration between nodes), artists seem to constantly collaborate within their communities with a few of them branching out into different communities. Rihanna seems to be playing a strong bridging role in her community exposing it to more genres and possibily making it easier for artists inside her community to do future collaborations through her existing ties.

Drake’s Pop Community

Rihanna seems to be connecting her own community to the electronic world through her colaborations with Calvin Harris, whom seems to be a very central figure in that group.

Clavin Harris Electronic Community

DJ Khalid Rap Community

By having a relationship with Calvin Harris she is inadvertedly also tied to this spesific cluster therefore resulting in Rihanna acting as a bridge. Consequently this can lead to other pop artists also starting to collaborate with artists in the electronic community.

She also seems to be extending her ties to the rap neighborhood through DJ Khalid.

Rihanna truly seems to Shine bright like a bridging node